Unverified50% confidenceFactExact time

数十乃至数百个决策树桩按权重叠加后，整体分类器的精度可以显著超越单棵深层决策树

1

Sources

50%

Confidence

Long-term

Relevance

10/4/2026

First Seen

Sources

从AdaBoost理解集成学习：弱模型如何组合成强模型

Related Entities

决策树桩concept AdaBoostconcept

Related Claims

Unverified在多步骤任务中持续自主决策的智能体，其错误行为比单次文本生成有更多机会被累积和放大，纠错窗口更狭窄74% similar Unverified贪心解码每一步都选择当前概率最高的词元，容易在多步任务中积累系统性偏差，因为局部最优不等于全局最优72% similar Unverified投机解码在草稿命中率较高时可获得数倍推理吞吐提升71% similar Unverified分类器的默认判定阈值会系统性地夸大语料库层面的显著性估计70% similar Unverified智能体面临误差级联问题：早期步骤判断误差会随迭代次数指数级放大70% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/973927

API

curl https://kongchang.com/api/v1/knowledge/claims/973927

MCP

get_claim(id=973927)